Efficiency Analysis of Production Systems

Abstract

The financial situation of transitional period in Russia makes companies in the country economize their resources, and as a consequence increase their efficiency. Data Envelopment Analysis (DEA) proposed by A. Charnes and W. Cooper is a powerful approach to determine the efficiency of the complex production systems[1,2,3]. At present, we witness a real scientific boom of DEA approach development, both theory and applications. Many scientific publications are devoted to DEA applications in the financial services sector. In our economic situation, the straight application of conventional DEA approach frequently leads to “strange” results. The analysis of these situations has driven us to the conclusion that we must introduce some additional constructions to the classical approach. In our work, we consider some new constructions in the DEA framework and apply them to efficiency analysis of the leading Russian banks and vertically integrated oil companies. The DEA approach generalizes many notions of macro- and microeconomics for the case of the multidimensional space of production parameters (inputs and outputs). However, it is very difficult for the manager to operate in the multidimensional space of parameters. A family of parametric optimization methods developed by our group enables us to make a cross-section of the efficient frontier by any pair of given directions. Thus, this will reduce the analysis of the complex system to the investigation of the well-known functions in economics: production function, isoquant, isocost, etc. This paper develops results that were presented at the 4th International Congress on Industrial and Applied Mathematics, Edinburgh, 1999 [4] and at the International DEA Symposium 2000, Brisbane [5].KeywordsData Envelopment AnalysisProduction UnitData Envelopment Analysis ModelEfficient FrontierData Envelopment Analysis ApproachThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.